Annals of Physical and Rehabilitation Medicine 65 (2022) 101594

Available online at

ScienceDirect
www.sciencedirect.com

Original article

Cognitive assessment scale for stroke patients (CASP): A multicentric

validation study
Charles Benaima,b,*, Gre
goire Wauquiezc, Dominic Pe

rennoud,e, Ce

line Piscicellid,e,
Brigitte Lucas-Pineau , Huei-Yune Bonnin-Koang , Philippe Vuadensb, Christine Binqueth,
f g

Abderrahmane Bourredjemh, Herve
Devilliersh

a
Physical Medicine and Rehabilitation, Lausanne University Hospital and University of Lausanne, Avenue Pierre Decker 4, CH-1011, Switzerland
b
Clinique Romande de Re
adaptation - SuvaCare, Sion, Switzerland
c
Physical Medicine and Rehabilitation, Dijon University Hospital, France
d
Neurorehabilitation Department, Institute of Rehabilitation, Grenoble Alpes University Hospital, Echirolles 38434, France
e
Laboratoire de Psychologie et NeuroCognition, UMR CNRS 5105, Universite
Grenoble Alpes, Grenoble, France
f
Centre de Re
e
ducation Fonctionnelle Divio, Dijon, France
g
Neurorehabilitation, N^ımes University Hospital, France
h
Centre d’Investigation Clinique CIC 1432 and University of Bourgogne-Franche-Comte
, France

A R T I C L E I N F O A B S T R A C T

Article History: Background: The Mini Mental State Examination and Montreal Cognitive Assessment are commonly used as
Received 16 December 2020 short screening batteries for assessing cognitive impairment after stroke. However, aphasia or hemispatial
Accepted 19 September 2021 neglect may interfere with the results. For this reason, we developed the Cognitive Assessment scale for
Stroke Patients (CASP), which takes these conditions into consideration and previously demonstrated its
Keywords: superiority over these scales in terms of feasibility.
Validation studies
Objectives: Our goal was to verify the psychometric properties of the (original) French version of the CASP.
Stroke
Methods: We included 201 patients with a recent first hemispheric stroke and 50 controls. Stroke patients
Cognitive
Assessment
were examined 4 times (visit 1 [V1] to visit 4 [V4]) in the subacute post-stroke phase. The structural validity
of the CASP was studied by principal factorial analysis, convergent validity by comparison with several varia-
bles including a comprehensive neuropsychological assessment, divergent validity by comparison with the
total score between stroke patients and controls, and sub-scores between right and left stroke. Internal con-
sistency, reproducibility and sensitivity to change were assessed. We propose the Minimal Clinically Impor-
tant Difference (MCID) value and a pathological threshold as well as a threshold to predict cognitive change
between V1 and V4.
Results: Of the 201 participants included (63% male; mean [SD] age 63 [13] years), CASP data were available
for 199/150/133/93 at V1/V2/V3/V4, respectively. CASP has a one-dimensional structure. The hypotheses of
convergent/divergent validities were confirmed. Internal consistency was good and reliability excellent.
Responsiveness was small to moderate, but the MCID could still be estimated. We discuss the choice of a
pathological threshold and a predictive threshold of V1 over V4.
Conclusions: CASP has good psychometric properties for screening cognitive impairment in the subacute
post-stroke phase, which is consistent with its Italian and Korean versions. It can be used for patients with
severe motor aphasia or left hemispatial neglect but not in case of severe oral comprehension or visual
impairment.
© 2021 The Author(s). Published by Elsevier Masson SAS. This is an open access article under the CC BY-NC-
ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abbreviations: ASRS, aphasia severity rating scale; AUC, area under the receiver oper-
ating characteristic curve; BDAE, boston diagnostic aphasia examination; CASP, cogni-
tive assessment scale for stroke patients; CFI, comparative fit index; CNA,
comprehensive neuropsychological assessment; MCID, minimal clinically important
difference; MMSE, mini mental state examination; MoCA, montreal cognitive assess-
ment; PFA, principal factor analysis; RMSEA, root mean square error of approximation;
SEM, standard error of measurement; SRM, standardized response mean; SRMR, stan-
dardized root mean square residual; TLI, tucker lewis index
* Corresponding author at: Physical Medicine and Rehabilitation, Lausanne Univer-
sity Hospital and University of Lausanne, Avenue Pierre Decker 4, CH-1011, Switzer-
land.
E-mail address: Charles.benaim@chuv.ch (H. Devilliers).

https://doi.org/10.1016/j.rehab.2021.101594
1877-0657/© 2021 The Author(s). Published by Elsevier Masson SAS. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/)
C. Benaim, G. Wauquiez, D. Pe
rennou et al.
Introduction
The Mini Mental State Examination (MMSE) [1] and Montreal
Cognitive Assessment (MoCA) [2] are 2 short screening batteries
assessing cognitive impairment that are widely used in neurology
but are not suitable for many post-stroke patients. For example, a
patient with motor aphasia cannot answer the MMSE and MoCA
questions assessing anterograde amnesia or time orientation. Simi-
larly, a patient with severe spatial neglect may be uncomfortable
with the MoCA naming item because it is visual (drawing of a lion on
the left side of the page, and so should be compensated by position-
ing the sheet to the right of the patient).
For this reason, we developed the Cognitive Assessment scale for
Stroke Patients (CASP) [3,4] to minimize these disadvantages (Sup-
plemental Material 1 and 2): all answers can be given without using
language (with the exception of the naming test, of course), and sev-
eral items are arranged on the test page so that that they are not hin-
dered by left spatial neglect (side most often affected). The CASP
contains 9 items grouped in 6 dimensions: communication, spatial/
visuo-construction, executive functions, short-term memory, praxis
and time orientation (6 £ 6 points). Its main features are that it
focuses specifically on post-stroke cognitive impairments, it is more
feasible to administer than the MMSE and MoCA in these patients,
and the severity of aphasia has much less influence on the overall
CASP score than the other 2 tests [3−5]. For example, it can be used
to establish that a mute patient is well oriented to time or does not
have anterograde amnesia, which is impossible to reveal with the
MMSE and MoCA. In Barnay et al., the CASP was impossible to admin-
ister in 18% of 44 unselected aphasic stroke patients as compared
with 36% for the MMSE and 30% for the MoCA [3]. In Benaim et al.,
the CASP was impossible to administer in 0% of 50 unselected non-
aphasic patients versus 0% for the MMSE and 6% for the MoCA [4]. A
cultural adaptation of the drawings was suggested by Park et al., who
cross-culturally validated the CASP in Korean [6], then by an Iranian
team that is currently validating the instrument in Persian (unpub-
lished data). Its main limitation is that it cannot be used in case of
severe oral comprehension impairment.
This work was the final validation of CASP (French version)
including structural validity, reliability and responsiveness. We
hypothesized that the CASP has a predominantly one-dimensional
structure, which corresponds to overall cognitive performance, and
possibly 2 additional dimensions supported by language and hemi-
spatial neglect, usually found in left and right hemispheric strokes,
respectively.
Material and methods
Study design and setting
Five rehabilitation services in France and Switzerland participated
in this multicenter cohort prospective study. The protocol was
approved by the ethics committees in France (2013-A00913-42, East-
France CCP, 2015-02-20) and Switzerland (CCVEM 048/14, 2014-
12.01).
Patients
In total, 201 consecutive patients aged 18 to 90 years were
recruited between January 2015 and February 2019. The inclusion
criteria were first-ever unilateral hemispheric stroke, time since
stroke < 2 months, and informed consent given. Non-inclusion crite-
ria were known cognitive deficiencies, psychosis or severe visual
impairment before the stroke, non-French-speaking, and severe
stroke-induced impairment of oral comprehension. Regarding the
latter and as explained previously [3,4], the assessment of oral com-
prehension by precise clinical tools such as the Boston Diagnostic
2
Annals of Physical and Rehabilitation Medicine 65 (2022) 101594
Aphasia Examination (BDAE) [7] items is not compatible with the
objective of the CASP, which must be able to be administered quickly
and by non-specialists. For this reason, we use the Aphasia Severity
Rating Scale (ASRS) of the BDAE to detect patients unable to complete
the CASP. However, the examiner is instructed to consider only oral
comprehension because severe speaking impairment does not pre-
clude taking the CASP. Under these conditions, the ASRS score should
not be ≤2/5.
Control groups
We included 50 age-matched patients recruited from orthopaedic
and geriatric rehabilitation services. Apart from stroke, the criteria
for inclusion and non-inclusion were the same as for stroke patients.
Data for these patients were used for assessing divergent validity. In
daily practice, the primary goal of the CASP is to detect cognitive
impairment in stroke patients. Therefore, we set up a second control
group of stroke patients without any cognitive impairment detected
by the neuropsychologist at the fourth visit. Data for these patients
were used for assessing pathological and predictive thresholds.
Variables and follow-up
After the inclusion visit (V0), patients were assessed for an initial
assessment (V1), then for 3 more visits (V2 to V4), according to the
schedule described in the Fig. 1.
Data collected included general socio-demographic data; brain
lesion description: side, mechanism (ischemia/hemorrhage), US
National Institute of Health Stroke Score7 (NIHSS); and neurological
deficiencies and impairments: CASP, MMSE, ASRS [8], modified Ran-
kin Scale [9] (mRankin), and the “sadness” item of the visual analogue
mood scale [10]. As a clinical anchor to define cognitive impairment,
we let the neuropsychologists choose among valid clinical scales
available in their respective rehabilitation centers the best-fitting
tests for their patients. However, we encouraged them to use the
BDAE or the Montreal-Toulouse 86 test [11] for aphasia and the Line
bisection [12] or Bell's test [13] for spatial neglect. Then, they had to
rate each of the following domains on a 5-point Likert scale (0: no
deficiency to 5: severe deficiency): language production and compre-
hension, praxis, amnesia, spatial unilateral neglect, executive func-
tioning and timed-orientation (same domains as for the CASP). We
called it the Comprehensive Neuropsychological Assessment (CNA).
All the neuropsychologists had at least 5 years of experience with
stroke.
On visits V1 to V4, the CASP was administered by experts (neuro-
psychologists, senior neurologists or physiatrists) or non-expert clini-
cians (physicians in training or clinical psychologists), depending on
the local organization. In all cases, the neuropsychologist who admin-
istered the CNA was blinded to CASP results. The sequence of exam-
iners for visits 1 to 3 (reliability) was “1st rater−2nd rater−1st rater”
in all but one center (1st rater−1st rater−2nd rater). All other clinical
scales were administered by physicians (physiatrists or neurologists).
V1 to V4 always occurred during the rehabilitation stay; patients
who left the hospital before V4 were not contacted after discharge.
Statistical methods
The statistical methods used follow the COSMIN guidelines, and
the reporting is in accordance with the STARD 2015 checklist.
Structural validity: The aim was to determine the underlying
dimensions of the CASP (i.e., 1, 2 or more concepts measured by the
items). First, we conducted principal factor analysis (PFA), which
builds up uncorrelated variables (factors). Coefficients defining these
linear combinations, called factor loadings, may be interpreted as cor-
relation coefficients. To determine the number of retained factors, we
used the Kaiser criterion [14]: eigenvalues (proxy of the explained
C. Benaim, G. Wauquiez, D. Pe
rennou et al.
•Inclusion - Informed consent
•ASRS, NIHSS, mRankin, VAMS
Vo •N=201
•D0 (Validity)
•CASP-V1, MMSE, CNA-1
V1 •N=193 (121 men, mean [SD] age 63 [13] years, mean NIHSS 5.0 [4.0])
•D2-4 (Intra-rater reliability)
•CASP-V2
V2 •N=151 (92 men, mean age 62 [13] years, mean NIHSS 5.4 [4.1])
•D5-7 (inter-rater reliability)
•CASP-V3
V3 •N=135 (80 men, mean age 62 [14] years, mean NIHSS 5.5 [4.0])
•D21-60 (Sensivity to change)
•CASP-V4, CNA-2, evoluon
V4 •N=93 (54 men, mean age 62 [12] years, mean NIHSS 6.0 [4.0])
Fig. 1. Scheduling and content of visits from the inclusion visit (V0) to visit 4 (V4), number of patients per visit. ASRS, Aphasia Severity Rating Scale; CASP, Cognitive Assessment
scale for Stroke Patients; CNA, Comprehensive Neuropsychological Assessment; D, day; MMSE, Mini Mental State Examination; mRankin, modified Rankin Scale; NIHSS, US National
Institute of Health Stroke Score; VAMS, Visual Analogous Mood Scale.
variance) >1. We next studied item loading on each factor, consider-
ing 0.3 as significant loading [15]. Finally we performed a confirma-
tory factor analysis [16] by using the retained PFA structure of the
CASP. In this analysis, each item was defined to represent only one
domain, but the domain scores were allowed to correlate with each
other. To assess the model quality of fit, we report the following: (1)
ratio of the chi-square statistic and degrees of freedom (chisq/df ≤3
indicates acceptable fit) [17]; (2) standardized root mean square
residual (SRMR ≤.08 indicates acceptable fit) [18]; (3) root mean
square error of approximation (RMSEA<0.05 or even 0.08 is generally
considered good fit and >0.1 corresponds to poor fit) [19,20]; and (4)
indexes to describe incremental fit, the Bentler Comparative Fit Index
(CFI) and Tucker Lewis Index (TLI) (CFI and TLI >0.9 indicate good fit)
[21].
External validity: (1) convergent validity was determined by com-
paring the results of the CASP with those of other clinical scores by
Spearman’s rank correlation. We expected at least moderate correla-
tion (>0.35) with the CNA, MMSE, and mRankin scores and NIHSS. (2)
Divergent validity: we compared CASP domain scores between stroke
groups defined by lesion side (left vs right) and between cases and
controls by Student t test [22]. Left stroke patients were expected to
have lower scores on language items and higher scores on spatial/
visuo-construction items. Controls were expected to have higher
total scores than stroke patients.
Reliability: (1) for internal consistency, Cronbach’s alpha coeffi-
cient was computed for CASP items, with ≥ 0.7 considered satisfac-
tory [23]; (2) inter- and intra-rater concordance were assessed by
computing intra-class coefficients (ICCs). Agreement was considered
excellent at ICC >0.75, good at 0.4−0.75 and poor otherwise [24].
Responsiveness: The sensitivity to change observed by neuropsy-
chologists between V1 and V4 was quantified on a 9-point Likert
scale (from -4 to 4). Effect size (ES=Dscore/Ϭinitial score) and the
standardized response mean (SRM=Dscore/Ϭ [Dscore]) were calcu-
lated for patients with improved condition (improvement of at least
2 points on the Likert scale). Values of 0.20, 0.50, and 0.80 for these
statistics correspond to a small, moderate or important change,
respectively [25]. The mean difference in scores between patients
with improvement of at least 2 points on the CNA and those with no
improvement was considered a first estimate of the CASP Minimal
Clinically Important Difference (MCID), according to the Anchor-
based method [26]. A second estimate of the MCID was obtained
3
Annals of Physical and Rehabilitation Medicine 65 (2022) 101594
with the distribution approach by calculating the standard error of
measurement (SEM=Ϭ*Square root [1−r], where Ϭ=baseline standard
deviation and r=Cronbach’s alpha) [27].
Pathological thresholds of the CASP were determined by compar-
ing patients without cognitive deficiencies (CNA=0) to the others. We
used V4 data to ensure sufficient patients without impairment. The
area under the receiver operating characteristic curve (AUC) was cal-
culated for the CASP. Because the purpose of the CASP was for screen-
ing cognitive deficiencies by non-expert clinicians before referral to
neuropsychologists, the thresholds were determined by prioritizing
sensitivity to specificity.
Predictive thresholds were determined in the same manner: we
estimated the extent to which the CASP score at V1 could predict the
CNA score at V4.
Study size
The number of patients required for the PFA was 20 times the
number of CASP items, so 180. In total, 120 patients were required to
ensure an ICC of 0.7 with accuracy of §0.1. Fifty control participants
were needed to ensure a difference of 1.3 for divergent validity, with
precision §1.5.
Results
Participant characteristics (Table 1, Fig. 1)
We included 201 patients. Because the number of participants
needed for reliability (V2 and V3) was reached before the end of the
inclusions, the last patients had only V0 (inclusion), V1 (validity) and
V4 (sensitivity to change). The number of patients assessed at the 5
visits was 201, 193, 151, 135 and 93. The number of missing CASP
scores at visits V1 to V4 was 2, 1, 2 and 0, or 5/572 (1%). In all cases,
only one item was not completed (4, “Reproducing a copy of a cube”
and 1, “Inhibition/Flexibility”). The proportion of incomplete MMSE
scores was 10/152 (7%). For participants who required a long stay in
rehabilitation services and who were still present at V4, the mean
(SD) CASP score improved by 1.87 (3) points (p < 0.001). At initial
assessment, they had more severe global impairments and disabil-
ities than the others (mean NIHSS: 5.84 [3.89] vs 4.01 [3.90],
p = 0.0017; mRankin score: median [interquartile range] 4 [3, 4] vs 2
C. Benaim, G. Wauquiez, D. Pe
rennou et al. Annals of Physical and Rehabilitation Medicine 65 (2022) 101594

Table 1 Table 2
Main baseline characteristics of stroke patients (n = 201). Factor loadings of the principal factor analysis. All items were best repre-
sented on the first factor.
Overall Patients attending V4
n = 201 n = 93 Factor pattern
CASP items Factor1 (81%) Factor2 (13%) Factor3 (6%)
Age, years 63 (13) 62 (12)
Sex (M) 127 (63%) 54 (58%) Naming 0.45154 0.39273 -0.00212
Lesion side (n = 199) Comprehension 0.53658 0.33151 0.12017
Right 106 (53%) 56 (60%) Cube 0.55542 -0.29031 0.11372
Left 92 (46%) 35 (38%) Line bissection 0.33905 -0.19757 0.26231
Bilateral 1 5 (1%) Graphic series 0.62828 -0.03099 -0.03026
Lesion type Inhibition/flexibility 0.59698 -0.12880 -0.25295
Ischemia 152 (76%) 64 (68%) Short-term memory 0.52860 0.06323 0.07957
Hemorrhage 49 (24%) 29 (32%) Praxis 0.58015 -0.09607 -0.06064
Days post-stroke (V0) 31.1 (12.7) 33.7 (13.7) Temporal orientation 0.54706 -0.01520 -0.09595
NIHSS (V0, n = 196) 5.0 (4.0) 6.0 (4.0)
ASRS (V0, n = 200) 5.0 (1.4) 5.1 (1.4)
MMSE (V1, n = 151) 25.2 (4.2) 24.5 (5.0)
CNA (V1, n = 188) 5.4 (3.9) 5.9 (4.2) +Comprehension) was significantly lower and spatial/visuo-construc-
mRankin tion score (Cube+Line bisection) higher with left than right hemi-
- No disability (0, 1) 47 (23%) 7 (8%) spheric stroke (5.5 [1.1] vs 5.8 [0.4], p = 0.0053 and 4.9 [1.4] vs 4.1
- Light-mild disability (2, 3) 70 (35%) 31 (33%)
- Severe disability (4, 5) 79 (39%) 53 (57%)
[1.7], p = 0.0006). The mean CASP score was significantly lower for
VAMS sadness (n = 178) 71.4 (26.8) 74.9 (26.7) stroke patients than controls (30.6 [4.9] vs. 32.9 [3.2], p < 10 4).
CASP Scores for all 6 CASP domains were lower for stroke patients than
V1 (n = 199) 30.6 (4.9) 30.1 (5.6) controls, but the difference was not significant for praxis and time
V2 (n = 150) 31.1 (4.9) 30.8 (5.3)
orientation (p = 0.1911 and p = 0.4556, respectively).
V3 (n = 133) 31.6 (4.5) 31.1 (5.0)
V4 (n = 93) 31.8 (4.6) 31.8 (4.6)
Reliability
Data are mean (SD) unless indicated. ASRS, Aphasia Severity Rating
Sacle; CASP, Cognitive Assessment scale for Stroke Patients; CNA, Com-
prehensive Neuropsychological Assessment; MMSE, Mini Mental State Internal consistency: Cronbach's alpha was 0.78 (95% CI 0.69
Examination; mRankin, modified Rankin Scale; NIHSS, US National −0.83), which indicates good internal consistency.
Institute of Health Stroke Score; V, visit; VAMS, Visual Analogous Mood Inter-rater reliability: the mean time between the 2 administra-
Scale.
tions was 2.8 (1.0) days. The ICC was excellent for the total CASP score
(0.78, 95% CI 0.73−0.83]) and the language domain (0.80, 95% CI 0.74
[1−3], p < 0.0001) but equivalent cognitive impairments (mean CNA −0.84]) and good for the other 5 domains (0.42−0.65). Among the 9
score: 5.9 [4.2] vs 4.9 [3.6], p = 0.1194; mean CASP score: 30.1 [5.6] vs individual items, only Line bisection had a poor ICC (0.37, 95% CI 0.27
31.1 [3.91], p = 0.4024). After checking and validation, data for 201 −0.48]).
participants (193 at V1 and 8 at V2) were available for assessing Intra-rater reliability: The mean time between the 2 administra-
structural and external divergent validity (initial CASP data), 193 for tions was 3.7 (1.4) days. The ICC was excellent for the total CASP score
external convergent validity (V1), 132 for inter-rater reliability (0.85, 95% CI 0.79−0.89) and language (0.89, 95% CI 0.85−0.92) and
(mainly V1 and V2), 129 for intra-rater reliability (mainly V2 and V3) spatial/visuo-construction domains (0.77, 95% CI 0.69−0.83) and
and 93 for sensitivity to change (V4). The mean age of the 50 non- good for the other 4 domains (0.51−0.73). For the 9 individual items,
stroke participants was comparable to that of stroke patients (64 [17] the ICCs were good or excellent (0.51−0.96).
vs 63 [13], p = 0.68) but with fewer men (42% vs 63%, p = 0.0097).
Responsiveness
Structural validity
If we consider as significantly improved individuals showing
PFA: Only one factor was selected by the Kaiser Criterion, which improvement by at least 2 points on the CNA (78/93 stroke patients),
suggests that CASP is rather unidimensional. Factor loadings of all then the responsiveness was small (ES=0.41) to moderate
items were higher for Factor 1 than other factors (Table 2), which (SRM=0.76). For these patients, CASP scores improved by a mean of
therefore can be considered the “overall cognitive performance fac- 2.3 (3.0) points (p = 0.002), which can be considered a first estimate
tor”. Factors 2 and 3 should not be taken into account, but of note, of the MCID (anchor-based method). The second estimate (distribu-
language items (left lesions) were best correlated with Factor 2 and tion approach) was very close: 2.3.
the hemispatial neglect item (right lesions) with Factor 3.
Confirmatory factor analysis: Chisq/df and SRMR values were Pathological threshold (Table 3)
acceptable (2.79 and 0.0643, respectively), the RMSEA was adequate
(0.095, 95% confidence interval [CI] 0.069−0.121), and CFI and TLI At V4, the mean CASP score for the 11 patients without cognitive
were fair (0.864 and 0.812). impairment (CNA score =0 at V4) was 35.6 (0.5) versus 31.2 (4.6) for
the other 82. The AUC was excellent: 0.90 (95% CI 0.84−0.96). The
External validity threshold was 34.5 to 35 depending on whether sensitivity or speci-
ficity was prioritized. A priori, we retained a threshold of 35, which
Convergent validity: the CASP score was strongly correlated with gave a sensitivity of 89% and specificity 64%. Positive and negative
the CNA score (Rho = -0.788, p < 10 4) and MMSE score (Rho = 0.640, predictive values to predict cognitive impairment were 95% and 56%,
p < 10 4), moderately with the mRankin score (Rho = -0.388, respectively.
p < 10 4) and weakly with the NIHSS (Rho = -0.288, p < 10 4); the
latter did not support the convergent validity. Predictive threshold (Table 4)
Divergent validity: Mean CASP scores were similar for left and
right hemispheric stroke patients (30.8 [5.3] and 30.3 [4.6], At V1, the mean CASP score for the 11 patients who fully recov-
p = 0.5001). As expected, the mean language score (Naming ered cognitive functions at V4 (CNA=0 at V4) was 34.2 (1.7) versus
4
C. Benaim, G. Wauquiez, D. Pe
rennou et al. Annals of Physical and Rehabilitation Medicine 65 (2022) 101594

Table 3 calculations, the number of stroke patients without cognitive

Pathological thresholds. Sensitivity, specificity and Youden index for the impairment was quite small as compared with other patients, and
CASP at visit 4 (CASP V4) as compared with the Comprehensive Neuropsy-
this value should be confirmed with a larger population. In contrast,
chological Assessment at visit 4 (CNA V4).
we prioritized the sensitivity of this screening test; otherwise the
CASP V4: Potential cut-off Sensitivity Specificity Youden Index threshold would have been lower. Several control participants had a
33.0 0.54 1.00 0.54 score below this value, which may be a surprise. We had included
33.5 0.56 1.00 0.56 these participants for the divergent validity check, and for this, they
34.0 0.74 1.00 0.74 did not need to undergo a full cognitive assessment before enrol-
34.5 0.76 1.00 0.76
ment. We simply ensured no known cognitive impairments in medi-
35.0 0.89 0.63 0.52
cal records. Thus, the CASP likely detected some participants as
having early-onset cognitive impairment that had not been previ-
29.5 (5.7) for the other 82. The AUC was very good: 0.81 (95% CI 0.70 ously explored. The threshold of 34/36 in the first month post-stroke
−0.92). The threshold was 32.5 to 34.5 depending on whether sensi- allows us to differentiate patients who will have an excellent
tivity or specificity was prioritized. A priori, we retained a threshold medium-term evolution (between V1 to V4) from other patients. This
of 34.5, which gave a sensitivity of 84% and specificity of 55%. Positive can be useful in establishing the prognosis and for planning manage-
and negative predictive values to predict full recovery were 93% and ment, as was pointed out with the Birmingham Cognitive Screen
32%, respectively. [28].

Interpretation
Discussion

The CASP is not free of flaws. The main one is certainly that it can-
Our study suggests that the CASP could be an interesting tool for
not be administered in the presence of severe comprehension defi-
screening patients with cognitive impairment in the subacute phase
ciencies because the patient may not understand instructions, but
of stroke. This finding is consistent with results from its Italian and
this limitation is common to most cognitive assessment tests. Fur-
Korean versions [5,6]. It could also give an idea of the cognitive evolu-
thermore, although to our knowledge, the CASP is the only test that
tion in the medium term.
offers a temporal orientation item that does not rely on language, we
In the present study, 7% of the MMSE scores could not be calcu-
did not find a non-verbal equivalent of the spatial orientation test
lated because of missing items versus only 1% for the CASP, which
(“What country/city are we in?”, “What floor of the building?” etc.).
shows the good feasibility of this test.
However, patients without aphasia could possibly be invited to give
The CASP is essentially unidimensional in structure and measures
an oral response to that item.
cognitive impairment overall. However, the language items also con-
The CASP is not the only clinical scale specifically designed for
tributed somewhat to the formation of a second factor and the
post-stroke. To the best of our knowledge, one of the oldest (2009) is
neglect item contributed somewhat to the formation of a third factor.
the Brief Neuropsychological Screening test [29], which is very rapid
This result was not surprising because patients with severe aphasia
(5−10 min) but does not explore temporal orientation and is only
and those with severe unilateral spatial neglect are known to have a
available in Italian. The Birmingham Cognitive Screen [30] was pro-
very different clinical profile.
posed in 2012 for assessing apraxia and determining functional prog-
We verified the convergent and divergent validity, reproducibility
nosis after stroke. Its usefulness for global cognitive assessment and
and responsiveness of the CASP. The reproducibility of some items
prognosis was demonstrated in 2015 [30]. Thus, the tool is interest-
could be perfected, especially Line bisection between 2 different
ing and comprehensive, but it takes about 1 h to complete, which we
examiners. This observation could be due in part to the heterogeneity
felt was too long for our purposes. Moreover, it is not free of charge.
of our examiners because in 52% of cases, one of the 2 examiners was
We found 5 other stroke-specific batteries in the literature, all pub-
not an experienced clinician (physician in training). However, we
lished in 2015 or later but containing a variable number of verbal
would need to improve the instructions for completing the tests for
items to assess functions other than language: the Brief Memory and
better reproducibility because this tool is intended for rapid screen-
Executive Test [31], the Oxford Cognitive Screen [32], the Mild Vascu-
ing “at the bedside”, and young doctors are often on the front line.
lar Cognitive Impairment assessment tool [33] and the Northwick
Consequently, the following sentence has been added to the instruc-
Park Examination of Cognition [34].
tion for the bisection test: “. . .the form must be placed exactly in front
In our opinion, the most interesting result of the previous CASP
of the patient and on a table without distractors.”
studies was that the severity of language impairment affected the
The determination of the MCID was beyond our expectations
CASP scores for items (other than language) much less than did the
because both estimation methods yielded almost identical values,
MMSE or MoCA (this last result having been established for the
which is relatively rare. The pathological threshold below which cog-
French and Italian versions of the CASP) [3,5]. However, the CASP
nitive impairment is likely was quite high, 35/36, which indicates
cannot be used with severe oral comprehension disorders, the limit
that the CASP is an “easy” scale. However, in this part of the
of feasibility being set to a score of <3/5 for the ASRS [8]. The face and
content validity of the French version of the CASP have been estab-
Table 4 lished [3,4]. The Italian and Korean versions of the CASP showed
Predictive thresholds. Sensitivity, specificity and Youden index of the CASP good psychometric properties [5,6]. Persian and Chinese versions are
at visit 1 (CASP V1) as compared with the CNA at visit 4 (CNA V4). being validated, after minor cultural adaptation (data not yet pub-
CASP V1 Potential cut-off Sensitivity Specificity Youden Index lished).

31.0 0.56 0.90 0.47

31.5 0.59 0.90 0.50
Study limitations
32.0 0.65 0.81 0.47
32.5 0.69 0.81 0.50 The first limitation of this study was the absence of a standardized
33.0 0.71 0.63 0.35 CNA across the 5 centers. However, we thought that imposing a sin-
33.5 0.72 0.63 0.36
gle assessment battery such as that proposed by Hachinski et al. [35]
34.0 0.80 0.54 0.34
34.5 0.83 0.54 0.38 on 5 university services would be less effective than letting experi-
enced neuropsychologists choose tests best suited to the patient's
5
C. Benaim, G. Wauquiez, D. Pe
rennou et al.
impairments, from among those they were used to administering.
There were some deviations from the protocol in our study. For
example, some patients were included beyond the planned 2 months
post-stroke (up to 90 days) and some assessments for reproducibility
were > 4 days apart (up to 6 days). These protocol deviations may
have underestimated the responsiveness and reproducibility of the
CASP, but the results were still satisfactory. Another limitation of the
study is the sample size, which did not guarantee the validation or
the determination of pathological thresholds in different age and
educational groups or in left/right stroke patients. This would justify
the continued inclusion of control and stroke patients. Finally, we
had 201 patients at V1 and 93 at V4, which could indicate attrition
bias. Actually, those who left the units before V4 were comparable to
others in terms of cognitive impairments but less affected physically
and therefore required a shorter stay.
Generalisability
From our work and clinical experience, the CASP would also be
useful in the chronic post-stroke phase, but we do not know if it
could also be used in the early phase. In an intensive care unit,
patients are typically lying down, which is already a feasibility issue
and requires adaptation. In addition, some patients would probably
not be able to be assessed by the CASP at one time because of fatiga-
bility.
Finally, this test should not be given more prominence than it was
designed for: rapid screening for post-stroke cognitive impairment
by less experienced examiners (e.g., young residents). For this reason,
it should be sensitive rather than specific. If the screening is positive,
the patient must be assessed more closely by a neuropsychologist or
speech-language pathologist to organize management. After a few
years of using this test, within this framework, the CASP would cer-
tainly have its place as an alternative to other rapid cognitive assess-
ment batteries.
Conclusions
These results are encouraging and show that the CASP is a useful
screening tool in daily practice in the subacute phase after stroke.
Patients with a pathological score should be referred to a neuro-
psychologist or speech-language therapist for full assessment and
treatment.
Supplemental material:
- English CASP (on-line)
- French CASP (on-line)
Declaration of Competing Interest
None declared.
Acknowledgments
We thank all the young and experienced physicians, therapists
and patients for their kind participation in this study.
Funding
French National “Projet Hospitalier de Recherche Clinique” (PHRC)
2012.
Supplementary materials
Supplementary material associated with this article can be found,
in the online version, at doi:10.1016/j.rehab.2021.101594.
6
Annals of Physical and Rehabilitation Medicine 65 (2022) 101594
References
[1] Folstein MF, Folstein SE, McHugh PR. Mini-mental state”. A practical method for
grading the cognitive state of patients for the clinician. J Psychiatr Res
1975;12:189–98. doi: 10.1016/0022-3956(75)90026-6.
[2] Nasreddine ZS, Phillips NA, Be
dirian V, Charbonneau S, Whitehead V, Collin I,
et al. The montreal cognitive assessment, MoCA: a brief screening tool for mild
cognitive impairment. J Am Geriatr Soc 2005;53:695–9. doi: 10.1111/j.1532-
5415.2005.53221.x.
[3] Barnay JL, Wauquiez G, Bonnin-Koang HY, Anquetil C, Pe
rennou D, Piscicelli C,
et al. Feasibility of the cognitive assessment scale for stroke patients (CASP) vs.
MMSE and MoCA in aphasic left hemispheric stroke patients. Ann Phys Rehabil
Med 2014;57:422–35. doi: 10.1016/j.rehab.2014.05.010.
[4] Benaim C, Barnay JL, Wauquiez G, Bonnin-Koang HY, Anquetil C, Pe
rennou D,
et al. The cognitive assessment scale for stroke patients (CASP) vs. MMSE and
MoCA in non-aphasic hemispheric stroke patients. Ann Phys Rehabil Med
2015;58:78–85. doi: 10.1016/j.rehab.2014.12.001.
[5] Crivelli D, Angelillo MT, Grippa E, Colucci A, Nardulli R, Balconi M. When is a novel
psychometric measure needed? A preliminary analysis regarding the cognitive
assessment for stroke patients (CASP) battery compared with MMSE and MoCA.
Appl Neuropsychol Adult 2018;25:410–6. doi: 10.1080/23279095.2017.1320556.
[6] Park KH, Lee HW, Park KB, Lee JY, Cho AR, Oh HM, et al. The Korean version of the
cognitive assessment scale for stroke patients (K-CASP): a reliability and validity
study. Ann Rehabil Med 2017;41:362–75. doi: 10.5535/arm.2017.41.3.362.
[7] Mazaux JM, Orgogozo JM. Échelle d’e
valuation de l’aphasie (BDAE). Editions Sci-
entifiques et Psychologiques, Issy-les-Moulineau 1982.
[8] Reinvang I, Graves R. A basic aphasia examination: description with discussion of
first results. Scand J Rehabil Med 1975;7:129–35.
[9] van Swieten JC, Koudstaal PJ, Visser MC, Schouten HJ, van Gijn J. Interobserver
agreement for the assessment of handicap in stroke patients. Stroke
1988;19:604–7. doi: 10.1161/01.str.19.5.604.
[10] Stern RA, Arruda JE, Hooper CR, Wolfner GD, Morey CE. Visual analogue mood
scales to measure internal mood state in neurologically impaired patients:
description and initial validity evidence. Aphasiology 1997;11:59–71. doi:
10.1080/02687039708248455.
[11] Nespoulos J., Lecours A., Lafond D., Lemay M., Puel M., Joanette Y., et al. Protocole
Montre
al-Toulouse d’examen linguistique de l’aphasie: MT-86 module standard

dition re
vise

e par Rene
e Be

land et Francine Giroux). 1992.
initial, M1b(2e e
[12] Schenkenberg T, Bradford DC, Ajax ET. Line bisection and unilateral visual neglect
in patients with neurologic impairment. Neurology 1980;30:509–17. doi:
10.1212/wnl.30.5.509.
[13] Gauthier L, Dehaut F, Joanette Y. The bells test: a quantitative and qualitative test
for visual neglect. Int J Clin Neuropsychol 1989;11:49–54.
[14] Kaiser HF. The application of electronic computers to factor analysis. Educ Psychol
Meas 2016. doi: 10.1177/001316446002000116.
[15] McElhone K, Abbott J, Teh LS. A review of health related quality of life in systemic
lupus erythematosus. Lupus 2006;15:633–43. doi: 10.1177/0961203306071710.
[16] Thompson B. Exploratory and confirmatory factor analysis: understanding concepts
and applications. Washington, DC: American Psychological Association; 2004.
[17] Jackson DL, Gillaspy JA, Purc-Stephenson R. Reporting practices in confirmatory
factor analysis: an overview and some recommendations. Psychol Methods
2009;14:6–23. doi: 10.1037/a0014694.
[18] Schreiber JB, Nora A, Stage FK, Barlow EA, King J. Reporting structural equation
modeling and confirmatory factor analysis results: a review. J Educ Res
2006;99:323–38. doi: 10.3200/JOER.99.6.323-338.
[19] MacCallum RC, Browne MW, Sugawara HM. Power analysis and determination of
sample size for covariance structure modeling. Psychol Methods 1996;1:130–49.
doi: 10.1037/1082-989X.1.2.130.
[20] Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis:
conventional criteria versus new alternatives. Struct Equ Model Multidiscip J
1999;6:1–55. doi: 10.1080/10705519909540118.
[21] Kline RB. Principles and practice of structural equation modeling. Guilford Publi-
cations; 2011.
[22] Rudner LM, Getson PR, Knight DL. Biased item detection techniques. J Educ Stat
1980;5:213–33. doi: 10.2307/1164965.
[23] Cronbach LJ. Coefficient alpha and the internal structure of tests. Psychometrika
1951;16:297–334. doi: 10.1007/BF02310555.
[24] Enderlein G, Fleiss JL. The design and analysis of clinical experiments Wiley, New
York − Chichester − Brislane − Toronto − Singapore 1986, 432 S., £38.35 Biom J
1988;30 304−304. doi: 10.1002/bimj.4710300308.
[25] Crosby RD, Kolotkin RL, Williams GR. Defining clinically meaningful change in
health-related quality of life. J Clin Epidemiol 2003;56:395–407. doi: 10.1016/
s0895-4356(03)00044-1.
[26] Make B. How can we assess outcomes of clinical trials: the MCID approach. COPD
2007;4:191–4. doi: 10.1080/15412550701471231.
[27] Wells G, Beaton D, Shea B, Boers M, Simon L, Strand V, et al. Minimal clinically
important differences: review of methods. J Rheumatol 2001;28:406–12.
[28] Bickerton WL, Riddoch MJ, Samson D, Balani AB, Mistry B, Humphreys GW. Sys-
tematic assessment of apraxia and functional predictions from the Birmingham
cognitive screen. J Neurol Neurosurg Psychiatry 2012;83:513–21. doi: 10.1136/
jnnp-2011-300968.
[29] Lunardelli A, Mengotti P, Pesavento A, Sverzut A, Zadini A. The brief neuropsycho-
logical screening (BNS): valuation of its clinical validity. Eur J Phys Rehabil Med
2009;45:85–91.
C. Benaim, G. Wauquiez, D. Pe
rennou et al.
[30] Bickerton WL, Demeyere N, Francis D, Kumar V, Remoundou M, Balani A, et al.
The BCoS cognitive profile screen: utility and predictive value for stroke. Neuro-
psychology 2015;29:638–48. doi: 10.1037/neu0000160.
[31] Brookes RL, Hollocks MJ, Khan U, Morris RG, Markus HS. The brief memory and
executive test (BMET) for detecting vascular cognitive impairment in small vessel
disease: a validation study. BMC Med 2015;13:51. doi: 10.1186/s12916-015-
0290-y.
[32] Demeyere N, Riddoch MJ, Slavkova ED, Bickerton WL, Humphreys GW. The oxford
cognitive screen (OCS): validation of a stroke-specific short cognitive screening
tool. Psychol Assess 2015;27:883–94. doi: 10.1037/pas0000082.
Annals of Physical and Rehabilitation Medicine 65 (2022) 101594
[33] Oh HS, Kim JS, Shim EB, Seo WS. Development and clinical validity of a mild vas-
cular cognitive impairment assessment tool for Korean stroke patients. Asian
Nurs Res 2015;9:226–34. doi: 10.1016/j.anr.2015.04.005.
[34] Williams PM, Johnson C, Swan S, Barber C, Murphy P, Devine J, et al. The north-
wick park examination of cognition: a brief cognitive assessment tool for use in
acute stroke services. Int J Ther Rehabil 2016;23:314–22.
[35] Hachinski V., et al. National institute of neurological disorders and stroke−Cana-
dian stroke network vascular cognitive impairment harmonization standards n.d.
10.1161/01.STR.0000237236.88823.47. 2020